Minimizing Live VM Migration Downtime Using OpenFlow Based Resiliency Mechanisms

Abstract

Live Virtual machine (VM) migration is a promising solution for data center (DC) administrators to achieve a wide range of objectives - from load balancing to disaster evacuation. Ideally, live VM migration is seamless and the challenge is to minimize the downtime during which the VM is not responding to requests or providing the service. The duration of the downtime is mainly influenced by the migration scheme utilized to transfer the VM states to its new location. In addition, a significant portion of the overall downtime is due to the update of the routing for all ongoing flows between clients and the migrating VM. Consequently, it is very important to resume network connectivity rapidly once the VM gets attached to the new location. In this paper, we propose several schemes that are based on Software Defined Networking (SDN) that allow a fast restoration of network connectivity for a migrating VM using different OpenFlow resiliency methods, which include FastFailover groups, packet bicasting and stateful forwarding. Our evaluation with a network emulator demonstrates the tradeoffs that our schemes achieve in terms of fast restoration of connectivity and effects due to temporal sub-optimal paths.